Spatial Interference Cancellation and Channel Estimation for Multiple-Input Multiple-Output Wireless Communication Systems

CHAPTER I

INTRODUCTION

Wireless communications is the rapidly growth segment of the communication industries. It has captured the attention of media and the imagination of the public, and become one of the necessary elements in the daily life. Since the establishment of cellular systems that experienced drastically growth over the last decade, they are now serving around two billion users worldwide. Indeed, cellular phones have become an important tool for both business and resident sectors in most developed countries, and are rapidly replacing wired systems in many developing countries. In addition, wireless local area networks (LANs) have been aggressively replacing wired networks in many homes, businesses, and campuses. From the past to the present, many new applications, including wireless sensor networks, automated highways and factories, smart home and appliances, and remote telemedicine, are emerging from research ideas to concrete systems. Further, the rapid growth of wireless system in conjunction with the rapid expansion of laptop and palmtop computers indicates a bright future for wireless networks, both as stand-alone system and as part of the larger networking infrastructure.

1.1 Wireless Communication Systems

There are two most common wireless communication systems involving in our everyday life: cellular telephone systems and wireless LANs. Therefore, it is of importance to introduce such systems.

1.1.1 Cellular Telephone Systems

Cellular telephone systems are popular and lucrative worldwide. Indeed, the wireless revolution is ignited by such systems. Cellular systems provide two-way voice and data communication with regional, national, or international coverage. Nowadays, these systems have evolved to support lightweight handheld mobile terminals operating inside and outside buildings at both pedestrian and vehicle speeds, in opposition to the old system where the terminals were installed inside vehicles with antennas mounted on the vehicle roof.

The basic principle of cellular systems is frequency reuse, which exploits the fact that signal power falls off with distance to reuse the same frequency spectrum at spatially separated locations. In fact, the coverage area of the cellular systems is divided into non-overlapping cells where some set of channels is assigned to each cell. This same channel set is used in another cell at some distance away. Operation within a cell is controlled by a centralized base station. The interference caused by users in different cells operating on the same channel set is called intercell interference. The spatial separation of cells that reuse the same channel set, i.e. the reuse distance, should be as small as possible so that frequencies are used as often as possible, thereby maximizing spectral efficiency. However, as the reuse distance decreases, intercell interference increases, due to the smaller propagation distance between interfering cells. Since intercell interference must remain below a given threshold for acceptable system performance, reuse distance cannot be reduced below some minimum value. In practice, it is quite difficult to determine this minimum value since both the transmitting and interfering signals experience random power variations due to the characteristics of wireless signal propagation. In order to determine the best reuse distance and base station placement, an accurate characterization of signal propagation within the cells is needed.

All base stations in a given geographical area are connected via a high-speed communications link to a mobile telephone switching office (MTSO), as shown in Fig.1.1. The MTSO serves as a central controller for the network, allocating channels within each cell, coordinating handoff between cells when a mobile traverses a cell boundary, and routing calls to and from mobile users. A new user located in a given cell requests a channel by sending a call request to the cell’s base station over a separate control channel. The request is relayed to the MTSO, which accepts the call request if a channel is available in that cell. If no channels are available, then the call request is rejected. A call handoff is initiated when the base station or the mobile in a given cell detects that the received signal power for that call is approaching a given minimum threshold. In this case, the base station informs the MTSO that the mobile requires a handoff, and the MTSO then queries surrounding base stations to determine if one of these stations can detect that mobile’s signal. If so, then the MTSO coordinates a handoff between the original base station and the new base station. If no channels are available in the cell with the new base station, then the handoff fails and the call is terminated. A call will also be dropped if the signal strength between a mobile and its base station drops below the minimum threshold needed for communication due to random signal variations.

Efficient cellular system designs are interference-limited, i.e. the interference dominates the noise floor since otherwise more users could be added to the system. As a result, any technique to reduce interference in cellular systems leads directly to an increase in system capacity and performance. Some methods for interference reduction in use today or proposed for the future systems include cell sectorization, directional and smart antennas, multiuser detection, and dynamic resource allocation.

The first generation of cellular systems used analog communications, since they were primarily designed in the 1960’s, before digital communications became prevalent. Second generation systems moved from analog to digital due to its many advantages. The components are cheaper, faster, smaller, and require less power. Digital systems also have higher capacity than analog systems since they can use more spectrally-efficient digital modulation and more efficient techniques to share the cellular spectrum. Due to their lower cost and higher efficiency, service providers used aggressive pricing tactics to encourage user migration from analog to digital systems, and today analog systems are primarily used in areas with no digital services. However, digital systems do not always work as well as the analog ones. Users can experience poor voice quality, frequent call dropping, and spotty coverage in certain areas. The third generation cellular systems are able to provide higher transmission rates than the second generation cellular systems. It still uses the digital modulation technique similar to the second generation one. In addition, it can provide different data rates depending on mobility and location. However, it is not compatible with the second generation systems, so service providers must invest in a new infrastructure before they can provide the third generation cellular system services. Until now, it still debates in many countries about the third generation standard as well as the worth of deploying such systems in comparison to upgrading the second generation systems to cope the demand for high data transmission services.

1.1.2 Wireless LANs

Wireless LANs provide high-speed data within a small region, e.g. a campus or small building, as users move from place to place. Typically, wireless devices that access these LANs are stationary or moving at pedestrian speeds. All wireless LANs standards in the U.S. operate in unlicensed frequency bands. The primary unlicensed bands are the industrial, scientific, and medical (ISM) bands at 900 MHz, 2.4 GHz, and 5.8 GHz, and the unlicensednational information infrastructure (U-NII) band at 5 GHz. In the ISM bands, unlicensed users are secondary users so must cope with interference from primary users when such users are active; meanwhile, there are no primary users in the U-NII band. A federal communications commission (FCC) license is not required to operate in these two bands. However, this advantage comes at the the price of additional interference caused by other unlicensed systems operating in these bands for the same reason. The interference problem can be minimized by setting a limit on the power per unit bandwidth for unlicensed systems. Wireless LANs can have either a star architecture, with wireless access points or hubs placed throughout the coverage region, or a peer-to-peer architecture, where the wireless terminals self-configure into a network.

The first generation wireless LANs were first proposed and designed in the early 1990’s, which were based on proprietary and incompatible protocols. Most operated within the 26 MHz spectrum of the 900 MHz ISM bands using direct sequence spread spectrum, with data rates on the order of 1-2 Mbps. Unfortunately, the lack of standardization for these products led to high development costs, low-volume production, and small markets for each individual product.

The second generation wireless LANs were further developed for improving a low data-rate in the first generation one. As such, the well-known IEEE 802.11b standard was proposed. It can support the data rates of around 1.6 Mbps (raw data rates of 11 Mbps), operating with 80 MHz of spectrum in the 2.4 GHz ISM bands by exploiting the direct sequence spread spectrum technology. The growth rate of 802.11b wireless LANs has explosively increased resulting from a relatively high volume of productions produced by many companies and the adoption of this technology in the computer industry, i.e. the integration of 802.11b wireless LANs cards in many laptop computers.

Two additional standards in the 802.11 family were developed to provide higher data rates than 802.11b: IEEE 802.11a and IEEE 802.11g. The 802.11a standard is based on multicarrier modulation, and can support 20-70 Mbps data rates. It operates with 300 MHz of spectrum in the 5 GHz U-NII band. In addition, it can accommodate a large number of users at higher data rates due to the larger bandwidth used in this standard, in comparison to the 802.11b standard. The other standard, 802.11g, also uses multicarrier modulation and can be used in either the 2.4 GHz and 5 GHz bands with speeds of up to 54 Mbps. Many wireless LANs cards and access points support all three standards to avoid incompatibility.

There are many technical issues to be addressed for wireless communication systems, including how to improve the system capacity with a high data rate and how to remove interference signals, e.g. spatial and temporal interference signals, from such systems. In the next section, a prominent system, namely a multiple-input multiple-output (MIMO) system, that can overcome such difficulties will be described.

1.2 Multiple Antennas and Space-Time Communications

In this section, the wireless communication systems with multiple antennas at the transmitter and receiver are presented. These systems are commonly known as MIMO communication systems. An exploitation of the multiple antennas at both transmitter and receiver can provide the performance advantages in terms of increasing data rates through multiplexing gain and improving error probability performance through diversity gain. In MIMO communication systems, the transmit and receive antennas can both be used for diversity gain. Multiplexing gain can be obtained by exploiting the structure of the channel gain matrix to create independent signalling paths that can be used to send independent data. Historically, the initial excitement about MIMO was started by the pioneering work of Winters [1], Foscini [2], Gans [3], and Telatar [4,5] predicting remarkable spectral efficiencies for wireless communication systems with multiple transmit and receive antennas. These spectral efficiency gains often require accurate knowledge of the channel at the receiver, and sometimes at the transmitter as well. Therefore, the issue about channel estimation is of crucially interest accordingly, especially for the systems employing coherent receivers or the systems employing transmit filters for compensating channel variations. In addition to spectral efficiency gains, inter symbol interference (ISI) and interference from other users can be reduced by using smart antenna techniques. The cost of the performance enhancements obtained through MIMO techniques comes at the expense of deploying multiple antennas, the extra space and power of these extra antennas, and the added complexity required for multi-dimensional signal processing. More details will be discussed as follows.

1.2.1 MIMO Wireless Fading Channels

Wireless fading channels of point-to-point MIMO communication systems with Lt- transmit and Lr-receive antennas can be classified into two categories: narrowband MIMO fading channels and frequency-selective MIMO fading channels. Narrowband MIMO fading channels are also called flat (or frequency-nonselective) MIMO fading channels. For the narrowband MIMO fading channels, there exists only a direct path connecting the Lt-transmit antennas onto Lr-receive antennas [6]. Therefore, there are LtLr channel links in such channel models. In addition, since there is no multiple paths in this kind of channels, the channels do not suffer from ISI. For the other channels, frequency-selective MIMO fading channels, there exists multiple paths in each pair of transmit and receive antennas. These channels could suffer from ISI when the bandwidth of such channels is larger relative to the channel’s multipath delay spread. There are two approaches to dealing with ISI in MIMO channels [6]. First, a channel equalizer can be used to mitigate the effects of ISI. However, the equalizer is much more complex in MIMO channels since the channel must be equalized over both space and time. Second, multicarrier modulation or orthogonal frequency division multiplexing (OFDM) can be employed as an alternative to equalization in frequency-selective fading channels. Frequency-selective MIMO fading channels exhibit diversity across space, time, and frequency, so ideally all three dimensions should be fully exploited in the signalling scheme.

Different assumptions can be assumed about the knowledge of the channels at the transmitter and receiver. For a static channel, the channel state information (CSI) at the receiver is typically assumed known, since the channel gains can be obtained easily by sending a training sequence for channel estimation. If a feedback path is available, then CSI at the receiver can be sent back to the transmitter to provide CSI at the transmitter. When the channel is not known at either the transmitter or receiver, then some distribution on the channel gain must be assumed. The most common model for this distribution is a zero-mean spatially white (ZMSW) model, where the channel gain is assumed to be identically independently distributed (i.i.d) zero-mean unit-variance, complex circularly symmetric Gaussian random variable. In general, different assumptions about CSI and about the distribution of the channel gain lead to different channel capacities and different approaches to space-time signalling.

1.2.2 Performance Advantages of MIMO Communication Systems

Different strategies for exploiting the multiple antennas at the transmitter and receiver lead to different performance advantages of the MIMO communication systems. Here, a summary of such advantages is presented.

1.2.2.1 MIMO Multiplexing Gain

When both the transmitter and receiver have multiple antennas, there is a mechanism for performance gain called multiplexing gain. The multiplexing gain of a MIMO system stems from the fact that a MIMO channel can be decomposed into a number R of parallel independent channels. By multiplexing independent data onto these independent channels, an R-fold increase in data rates can be achieved in comparison to a system with just one antenna at the transmitter and receiver. This increased data rate is called the multiplexing gain.

In order to obtain independent channels from the MIMO system, the MIMO channel with Lr ×Lt channel gains needs to be assumed known to both the transmitter and receiver. Let H be an Lr × Lt channel gain matrix and RH be a rank of H of the MIMO system. The process for obtaining the independent channels are summarized as follows. First, the singular value decomposition (SVD) is performed for H, to arrive at

1.2.2.2 MIMO Diversity Gain

Alternatively, the multiple antennas at the transmitter and receiver can be used to obtain diversity gain instead of capacity gain. In this setting, the same symbol, weighted by a complex scale factor, is sent over each transmit antenna, so that the input covariance matrix has unit rank. By pursuing this strategy, the error probability performance will be improved, since more channels are used for sending the same symbol, proportionally to a number of the transmit and receive antennas used. This improved error probability performance is called the diversity gain. This strategy corresponds to the transmit precoding and receiver shaping described in section 1.2.2.1 being just column vectors: V = v and U = u. This strategy provides diversity gain by coherently combining of the multiple signal paths. Channel knowledge at the receiver is typically assumed since this is required for coherent combining. The diversity gain then depends on the availability of the knowledge of channels at the transmitter. When the channel matrix H is known, the received signal-to-noise ratio (SNR) is optimized by choosing u and v as the principal left and right singular vectors of the channel matrix H, respectively. When the channel is not known, at the transmitter, the transmit antenna weights are all equal. In addition, the lack of transmitter CSI will result in a lower SNR and capacity than with optimal transmit weighting. This strategy has obviously reduces capacity relative to optimizing the transmit precoding and receiver shaping matrices at an additional benefit of the reduced demodulation complexity.

1.2.2.3 Multiplexing/Diversity Tradeoffs

From the previous sections, there are two mechanisms for utilizing multiple antennas to improve wireless system performance. One option is to obtain capacity gain by decomposing the MIMO channel into parallel channels and multiplexing different data streams onto these channels. This capacity gain is referred to as a multiplexing gain. However, the SNR associated with each of these channels depends on the singular values of the channel matrix. Alternatively, the strategy for exploiting the multiple antennas to achieve a diversity gain can be done by coherently combining the channel gains. It is not necessary to use the antennas purely for multiplexing or diversity. Some of the space-time dimensions can be used for diversity gain, and the remaining dimensions used for multiplexing gain.

The multiplexing/diversity tradeoff or, more generally, the tradeoff between data rates, probability of error, and complexity for MIMO systems has been studied in the literature, from both a theoretical perspective and in terms of practice space-time code designs [7–9]. These works have primarily focused on block fading channels with receiver CSI only since when both transmitter and receiver know the channels, the tradeoff is relatively straightforward: antenna subsets can first be grouped for diversity gain and then the multiplexing gain corresponds to the new channels with reduced dimension due to the grouping. For the block fading model with receiver CSI only, as the block length grows asymptotically large, full diversity gain and full multiplexing gain (in terms of capacity with outage) can be obtained simultaneously with reasonable complexity by encoding diagonally across antennas [10, 11]. For finite block lengths, it is not possible to achieve full diversity and full multiplexing gains simultaneously, in which case there is a tradeoff between these gains. It is well-known that, in the MIMO system with Lt-transmit and Lr-receive antennas, if all transmit and receive antennas are used for diversity, the error probability will be proportional to SNRLtLr. Moreover, some of these antennas can be used to increase data rates at the expense of diversity gain. It is also possible to adapt the diversity and multiplexing gains relative to channel conditions. Specifically, in poor channel states, more antennas can be used for diversity gain; whereas, in good states, more antennas can be used for multiplexing gain.

1.2.2.4 Smart Antennas

From the previous sections, the multiple antennas at the transmitter and/or receiver can provide diversity gain as well as increased data rates through multiplexing gain. Alternatively, sectorization or phased array techniques can be used to provide directional antenna gain at the transmit or receive antenna array. This directionality can increase the signalling range, reduce delay-spread (or, equivalently, ISI) and flat-fading, and suppress interference between users. In particular, interference typically arrives at the receivers from different directions. Thus, directional antennas can exploit these differences to null or attenuate interference arriving from given directions, thereby increasing system capacity. The reflected multipath components of the transmitted signal also arrive at the receiver from different directions, and can also be attenuated, thereby reducing ISI and flat-fading. The benefits of directionality that can be obtained with multiple antennas must be weighted against their potential diversity or multiplexing benefits, giving rise to multiplexing/diversity/directionality tradeoffs. Whether it is best to use the multiple antennas to increase data rates through multiplexing, increase robustness to fading through diversity, or reduce ISI and interference through directionality is a complex tradeoff decision that depends on the overall system design.

The most common directive antennas are sectorized or phased (directional) antenna arrays, and the radiation patterns for these antennas along with an omnidirectional antenna radiation pattern are shown in Fig.1.2.

Directional antennas (or smart antennas) typically use antenna arrays coupled with phased array techniques to provide directional gain, which can be tightly controlled with sufficiently many antenna elements. Phased array techniques work by adapting the phase of each antenna element in the array, which changes the angular locations of the antenna beams (angles with large gain) and nulls (angles with small gain), via the use of weight vector, as

From Fig.1.3, the smart antennas basically consist of three major parts: an array antenna, an array processor, and a demodulator. The most important part for the smart antennas is the array processor, in which the adaptive algorithm module plays a major role in controlling the weight vector w. The adaptive algorithms for adjusting the weight vector can be divided into two approaches: a non-blind approach, in which the training signal is used to provide a prior knowledge of the location of the desired signal, and a blind approach, in which no training signal is employed. For the non-blind adaptive algorithm approach, many methods had been proposed, e.g. the minimum mean square error (MMSE), the maximum signal-to-noise ratio (Max SNR), the linear constraint minimum variance distortionless response (LCMV) methods [12]. For the blind approach, the well-known constant modulus algorithm (CMA) is the most popular method. More details about the basic operation of the smart antennas can be found in [12]. The performance advantage of the blind approach is the enhanced bandwidth efficiency since no training signal is transmitted. However, this benefit comes at the expense of an additional complexity as well as the performance loss, in some cases.

1.2.3 Space-Time Modulation and Coding

Beside the MIMO communication systems, space-time modulation and coding are of interest and importance because they can provide an effective signalling and codes that can achieve a full diversity gain. Since the signal design extends over both space (via the multiple antennas) and time (via multiple symbol times), it is typically referred to as a space-time code. Most space-time codes are designed for quasi-static channels where the channel is constant over a block of certain symbol times, and the channel is assumed unknown at the transmitter.

As previously mentioned, it can be seen that the MIMO communication systems provide a significant performance enhancement to the wireless communications, including the increased rates through multiplexing gain, the enhanced error probability through diversity gain, and the cancellation of ISI and interference through smart antennas. At the first step, the MIMO communication systems were developed for the flat fading channels, whereby no ISI occurs. In practice, most channels are frequency selective, therefore, such systems could suffer from ISI. Furthermore, this ISI severely affects the performances of the MIMO communication systems. Hence, an equalizer in time domain is needed to mitigate such problem. However, the design of equalizer for the MIMO communication systems is quite complicated due to the nature of a multi-dimensional signal processing inherent in these systems. Alternatively, the multicarrier modulation technique is quite promising for mitigating such problem due to its good performance and simplicity for implementation. This technique can be directly applied to the MIMO communication systems, known as the MIMO-OFDM communication systems. In the next text, the multicarrier modulation is presented.

1.3 Multicarrier Modulation

The basic idea of multicarrier modulation is to divide the transmitted bitstream into many different substreams and send these over many different subchannels. Typically, the subchannels are orthogonal under ideal propagation conditions. The data rate on each of the subchannels is much less than the total data rate, and the corresponding subchannel bandwidth is much less than the total system bandwidth. The number of substreams is chosen to insure that each subchannel has a bandwidth less than the coherence bandwidth of the channel, so the subchannels experience relatively flat fading. Thus, the ISI on each subchannel is small. The subchannels in multicarrier modulation need not be contiguous, so a large continuous block of spectrum is not needed for high rate multicarrier communications. Moreover, multicarrier modulation is efficiently implemented digitally. In this discrete implementation, i.e. an OFDM system, the ISI can be completely eliminated through the use of a cyclic prefix.

Multicarrier modulation is currently used in many wireless systems. However, it is not a new technique; it was first used for military high frequency (HF) radios in the late 1950’s and early 1960’s. Starting around 1990, multicarrier modulation has been adopted in many diverse wired and wireless applications, including digital audio and video broadcasting in Europe, digital subscriber lines (DSL) using discrete multitone, and the most recent generation of wireless LANs. Multicarrier modulation is also a candidate for the air interface in next generation cellular systems. Multicarrier techniques are common in high data rate wireless systems with moderate to large delay spread, as they have significant advantages over time-domain equalization. In particular, the number of taps required for an equalizer with good performance in a high data rate system is typically large. Thus, these equalizers are highly complex. Moreover, it is difficult to maintain accurate weights for a large number of equalizer taps in a rapidly varying channel. For these reasons, most emerging high rate wireless systems use either multicarrier modulation or spread spectrum instead of equalization to compensate for ISI.

1.3.1 Data Transmission using Multiple Carriers

The simplest form of multicarrier modulation divides the data stream into multiple substreams to be transmitted over different orthogonal subchannels centered at different subcarrier frequencies. The number of substreams is chosen to make the symbol time on each substream much greater than the delay spread of the channel or, equivalently, to make the substream bandwidth less than the channel coherence bandwidth.

The subchannels can be divided to be non-overlapping subchannels. However, this strategy is spectrally inefficient, and near-ideal low pass filters is required to maintain the orthogonality of the subcarriers at the receivers. Moreover, it requires N independent modulators and demodulators, which entails significant expense, size, and power consumption. On the other hand, the subchannels can be divided to be overlapping subchannels, which make use of the spectrum more efficiently. It can be shown that the minimum frequency separation required for subcarriers to remain orthogonal over the symbol interval [0, TN] is 1/TN [6]. Typically, a set of sinusoidal function in conjunction with an appropriate baseband pulse shape is chosen to form a set of (approximately) orthonormal basis functions. Given this orthonormal basis set, even if the subchannels overlap, the modulated signals transmitted in each subchannel can be separated out in the receiver. In Fig.1.4, an illustration of the frequency-domain multicarrier with overlapping subcarriers is depicted. Note that f0, . . . , fN−1 denote subcarriers of subchannels, respectively.

1.3.2 Discrete Implementation of Multicarrier

Although multicarrier modulation was invented in the 1950’s, its requirement for separate modulators and demodulators on each subchannel was far too complex for most system implementations at the time. However, the development of simple and cheap implementations of the discrete Fourier transform (DFT) and the inverse DFT (IDFT) twenty years later, combined with the realization that multicarrier modulation can be implemented with these algorithms, ignited its widespread use.

1.3.2.2 Orthogonal Frequency Division Multiplexing

The OFDM implementation of multicarrier modulation, including a transmitter and receiver, is shown in Fig.1.5 and 1.6, respectively. The input data stream is modulated by a QAM modulator, resulting in a complex symbol stream X[0], . . . ,X[N − 1]. This symbol stream is passed through a serial-to-parallel convertor, whose output is a set of N parallel QAM symbols X[0], . . . ,X[N − 1] corresponding to the symbols transmitted over each of the subcarriers. Thus, the N symbol output from the serial-to-parallel convertor are the

1.4 Channel Estimation

For the wireless communication systems employing a coherent receiver, an accurate CSI is crucially needed. Thus, the issue of channel estimation is of most interest since the capacity of such systems as well as the error probability performance depends on this CSI estimate. An alternative way to cross over the channel estimation problem is to employ the differential modulation technique. However, this benefit of not explicitly performing the channel estimation comes at the expense of the 3-dB loss in a received SNR. In this section, an overview of channel estimation is presented. More specific details and approaches for channel estimation will be described in Chapter 3 and 4.

For the SISO communication systems, there are two ways to perform the channel estimation: blind channel estimation and non-blind channel estimation. Blind channel estimation doest not need any training (or pilot) signals for using as a prior information for channel estimation. Specifically, it actually exploits a special structure of the transmitted and received signals or the characteristics of channels, such as signal and noise subspaces of the received signal, a constant modulus property of the transmitted signal, and a cyclo-stationary property of the channels [15], for estimating the channels. The explicit benefit of the blind channel estimation is its enhanced bandwidth efficiency, since there is no bandwidth efficiency loss caused by a transmission of training signals. However, the computational complexity of the blind channel estimation is a prohibitive cost. Moreover, when the dimensions of signal processing increases, e.g. as in the MIMO systems, the computational complexity of such channel estimation increases and cannot be affordable. On the other hand, non-blind channel estimation can be achieved by the use of the training signal. Despite the bandwidth efficiency loss, the non-blind channel estimation has several advantages. First, its computational complexity is not that high, in comparison to that of blind channel estimation, and can be affordable even for the MIMO systems. Second, its performance in terms of a mean square error (MSE) is excellent, mostly better than the that of blind channel estimation. Hence, the non-blind channel estimation is of particular interest. There are two approaches for designing the training signal: pilot-symbol assisted modulation (PSAM) approach [16] and pilot-embedding approach [17].

In the PSAM approach, the training signal is time-multiplexing onto the transmit data stream. At the receiver, this training signal is extracted from the received signal, and then is used for channel estimation. The interpolation technique can be adopted for improving an accuracy of the channel estimate. In summary, an extra time-slot is needed for sending this training signal, resulting in the bandwidth efficiency loss. In addition, this loss is proportional to the amount of the training signal used.

Alternatively, in the pilot-embedding approach, sequences of the training and data are added up together to form the transmit data stream. At the receiver, such soft estimation algorithms, e.g. Viterbi’s algorithm, are used for recovering such training signal, and then, estimating the channels. The benefit of the pilot-embedding approach is the enhanced bandwidth efficiency, however, at the price of an increased computational complexity for channel estimation. Moreover, since these sequences are added up together, for a given power constraint, some power needs to be dedicated to the training. Hence, the power, i.e. the remaining power, dedicated to the data is then being reduced, resulting in an increasing error probability in the system.

There are many approaches for channel estimation, including a least square (LS) approach, an ML approach, and a linear minimum MSE (LMMSE) approach [18]. For the communication systems corrupted by the additive white Gaussian noise, the LS and ML channel estimation approaches behave similarly, and so does their performances [18]. For the LMMSE channel estimation, a prior information about the channel correlation is exploited; as a result, its performance is the best [18]. However, this performance enhancement comes at the expense of the increased computational complexity of the estimation process. In addition, the approaches proposed for SISO systems can be well applied to the MIMO systems.

1.5 Multiuser Systems

In multiuser systems, the system resource must be divided among multiple users. It is well-know that signals of bandwidth B and time duration T occupy a signal space of dimension 2BT. In order to support multiple users, the signal space of dimension 2BT of a multiuser system must be allocated to the different users. Allocation of signalling dimensions to specific user is called multiple access. Multiple access methods perform differently in different multiuser channels, and these methods will be applied to the two basic multiuser channels: downlink and uplink channels.

A multiuser channel refers to any channel that must be shared among multiple users. There are two different types of multiuser channels: the downlink and uplink channels. A downlink channel, also called a broadcast channel or forward channel, has one transmitter sending to many receivers. Since the signals transmitted to all users originate from the downlink transmitter, the transmitted signal is the sum of signals transmitted to all K users. Thus, the total signalling dimensions and power of the transmitted signal must be divided among the different users. Another important characteristic of the downlink is that both signal and interference are distorted by the same channel. An uplink channel, also called a multiple access channel or reverse channel, has many transmitters sending signals to one receiver, where each signal must be within the total system bandwidth B. In addition, the signals of different users in the uplink travel through different channels, the received powers associated with the different users will be different if their channel gains are different.

1.5.1 Multiple Access

Efficient allocation of signalling dimensions between users is a key design aspect of both uplink and downlink channels, since bandwidth is usually scarce. When dedicated channels are allocated to users, it is often called multiple access. Multiple access techniques divide up the total signalling dimensions into channels and then assign these channels to different users. The most common methods to divide up the signal space are along the time, frequency, and code axes. The different user channels are then created by an orthogonal or non- orthogonal division along these axes: frequency-division multiple access (FDMA) and time-division multiple access (TDMA) are orthogonal channelization methods whereas code-division multiple access (CDMA) can be orthogonal or non- orthogonal, depending on the code design.

1.5.1.1 Frequency-Division Multiple Access

In FDMA, the system signalling dimensions are divided along the frequency axis into nonoverlapping channels, and each user is assigned a different frequency channel, as shown in Fig.1.7. FDMA is the most common multiple access option for analog communication systems, where transmission is continuous. Multiple access in OFDM systems, called orthogonal frequency-division multiple access (OFDMA), implements FDMA by assigning different subcarriers to different users.

1.5.1.2 Time-Division Multiple Access

In TDMA, the system dimensions are divided along the time axis into non overlapping channels, and each user is assigned a different cyclically-repeating time slot, as shown in Fig.1.8. These TDMA channels occupy the entire system bandwidth, which is typically wideband, so some form of ISI mitigation is required. A major difficulty of TDMA is the requirement for synchronization among different users in the uplink channels. To maintain orthogonal time slots in the received signals, the different uplink transmitters must synchronize such that after transmission through their respective channels, the received signals are orthogonal in time. Multipath can also destroy time-division orthogonality in both downlinks and uplinks if the multipath delays are a significant fraction of a time slot. TDMA is used in many digital cellular phone standards.

1.5.1.3 Code-Division Multiple Access

In CDMA, the information signals of different users are modulated by orthogonal or non-orthogonal spreading codes. The resulting spread signals simultaneously occupy the same time and bandwidth, as shown in Fig.1.9. The receiver uses the spreading code structure to separate out the different users. CDMA is used for multiple access in IS-95 digital cellular standards, with orthogonal spreading codes on the downlink and a combination of orthogonal and non-orthogonal codes on the uplink. It is also used in the wideband CDMA (W-CDMA) and CDMA 2000 digital cellular standards.

1.6 Motivations and Scope of This Dissertation

In the future wireless communications, high data rate transmission services are highly demanded. As mentioned earlier, one of the promising communication schemes making such demand a reality is the MIMO communication system. In addition, several gains including multiplexing, diversity, and directionality gains inherent in the MIMO communication systems can significantly improve the system performances, e.g. the system capacity, the error probability, and the robustness to the interference. These performance advantages are the motivations for this dissertation. In this dissertation, the MIMO communication systems are considered. This dissertation studies the MIMO communication systems in two aspects: smart antenna system (or spatial interference cancellation scheme) and channel estimation.

It is well-known that the multiple access interference (MAI) can severely affect the performance of a CDMA system. Typically, the received signal consists of both the desired and the interference signals coming from different directions of arrival (DOA). Hence, the smart antenna system, i.e. a spatial filter, can be used to mitigate such interference by rejecting it as well as maintaining the desired signal in the desired direction. The advantages by employing such a system are the capacity improvement, the signal link enhancement, and the enhanced error probability. In the first part of this dissertation, the smart antenna system (or the spatial interference cancellation scheme), namely an interference-rejected blind array processing (IRBAP), is proposed for interference canceling receivers in direct sequence (DS) CDMA systems. IRBAP only exploits the spreading codes of users as the information for its weight adjustment procedures. This proposed scheme is robust to the closely-separated-DOA in-beam interference signals, especially in the near-far effect situation. The theoretical analysis, including convergence analysis and error probability analysis, for IRBAP is conducted. Performance evaluation via computer simulations is also performed.

For the wireless communication systems employing coherent receivers, an accurate channel estimate is crucially needed. This demand motivates the studies in the second part of this dissertation. In the second part of this dissertation, channel estimation for the MIMO communication systems is investigated.

Firstly, the channel estimation for the MIMO communication systems with flat fading channels is considered. The novel pilot-embedding technique, called a data-bearing approach for pilot-embedding for joint channel estimation and data detection, is proposed by exploiting the null-space property and the orthogonality property of the data-bearer and pilot matrices. The unconstrained ML and LMMSE channel estimators are proposed. The ML data detection is also studied. MSE of channel estimation, Cramer-Rao lower bound (CRLB), and Chernoff’s bound of bit error rate (BER) for space-time (ST) codes are analyzed for examining the performance of the proposed scheme. The optimum power allocation scheme for data and pilot parts is also investigated. Three data-bearer and pilot structures, including time multiplexing (TM)-based, ST-block code (STBC)-based, and code-multiplexing (CM)-based, are proposed. Computer simulations are conducted for evaluating the performance of the proposed scheme.

Secondly, the channel estimation for the MIMO-OFDM communication systems with frequency-selective fading channels is considered. Since the multipath delay profile of such channels is arbitrary in such system, an effective channel estimator is needed. The generalization of a data-bearing approach for pilot-embedding to such system is proposed. The pilot-embedded data-bearing (PEDB) LS channel estimation and ML data detection are considered. Then, an LS FFT-based channel estimator is proposed to improve the performance of the PEDB-LS channel estimator. The effect of model mismatch error inherent in the proposed LS FFT-based channel estimator when considering non-integer multipath delay profiles, and the performance analysis for such estimators are investigated, Under the framework of pilot embedding, an adaptive LS FFT-based channel estimator is proposed to improve the performance of that of LS FFT-based channel estimator. The optimum number of taps for this estimator is determined. Computer simulations are conducted for examining the performance of the proposed channel estimators.

In addition, the data-bearing approach for pilot embedding can be directly applied to the SISO communication system as well.

1.7 Outline

In chapter 2, the smart antenna system, i.e. IRBAP, is proposed for interference canceling receivers in DS-CDMA systems with flat-fading channels. In addition, the basic background for the interference canceling receivers and the basic blind beam forming is reviewed. Theoretical analysis, including convergence analysis and error probability analysis, for IRBAP is conducted. Performance evaluation via computer simulations is also performed. The results are summarized and discussed in the end of this chapter.

In chapter 3, a data-bearing approach for pilot-embedding frameworks for joint channel estimation and data detection in space-time coded MIMO systems with flat fading channels is proposed. The unconstrained ML and LMMSE channel estimators, and the ML data detection, are also proposed. MSE of channel estimation, CRLB, and Chernoff’s bound of BER for ST codes are analyzed. The optimum power allocation scheme for data and pilot parts is also investigated. Three data-bearer and pilot structures, including TM-based, STBC based, and CM-based, are proposed. Computer simulations are conducted for evaluating the performance of the proposed scheme. The results are summarized and discussed in the end of this chapter.

In chapter 4, the generalization of the data-bearing approach for pilot-embedding to the space-frequency coded MIMO-OFDM systems with frequency-selective fading channels is proposed. The PEDB-LS, LS FFT-based, and adaptive LS FFT-based channel estimators are proposed. In addition, the PEDB-ML data detection is also considered. The performance analysis for such channel estimators is conducted. Computer simulations are performed for examining the performance of the proposed channel estimators. The results are summarized and discussed in the end of this chapter.

In chapter 5, the dissertation is concluded, and the contributions of this dissertation as well as the future works are also mentioned.

 

CHAPTER II

AN INTERFERENCE-REJECTED BLIND ARRAY PROCESSING FOR INTERFERENCE CANCELING RECEIVERS IN CDMA SYSTEMS

This chapter presents the smart antenna system (or the spatial interference cancellation scheme) for DS-CDMA systems using a blind array processing, i.e. a blind beam forming. The basic background for the blind beam forming schemes in DS-CDMA systems are reviewed. Then the proposed smart antenna system, i.e. IRBAP, is presented. Furthermore, the performance analysis for IRBAP including convergence analysis, bit error rate analysis, and complexity comparison is also provided. For illustrating the performance of IRBAP, the simulation results in comparison with the analytical results are shown. In addition, the discussion and concluding remark are given in the end of this chapter.

2.1 Introduction

Many interference cancellation techniques have recently been proposed for wireless communications [19–26], where the smart antenna system or the beam forming is one of the most attractive and effective techniques. This technique plays a role of spatial filtering, where the received signal can be seen as a combination of users’ plane waves impinging on an array antenna from different directions; as a result, the desired signal can be captured as well as the interference signals can be rejected by the selectivity and nulling capabilities of an array antenna [27].

The most important component of the smart antenna system is the adaptive algorithm used for adjusting the weight vectors. Different algorithms can be categorized as non-blind adaptive algorithms, where the receiver uses a pilot signal sent by the transmitter for its weight adjustment, and blind adaptive algorithms, where the receiver adjusts its weight vectors without using the pilot signal. The advantage of non-blind adaptive algorithm is low computational complexity; however, this benefit comes at the price of channel bandwidth efficiency [12]. Consequently, the blind algorithm is more attractive than the non-blind adaptive algorithm for the benefit of no need of pilot signals, and hence, the bandwidth efficiency will be enhanced. Of particular interest in this chapter is to design the effective blind adaptive beam forming with an achievable low complexity.

One main class of blind adaptive algorithms in the literatures [24, 28–31] exploits the constant modulus property of the received signals. The idea is to capture the strongest constant modulus signal first and then subtract this captured signal from the overall received signals. Then in the next stage it will continue to capture the remaining signals and operate successively until the weakest signal is captured. This idea can be applied for multiuser detection purpose as well [32]. In a DS-CDMA system, several blind adaptive algorithms have been proposed, such as the Decision-Directed (DD) method, the CMA method, and the Despread-Respread (DR) method along with the LS technique [12, 33, 34]. These algorithms work well in the situation where power control is perfect and DOAs are well separated. However, in the near-far effect situations with closely separated DOAs, these algorithms fail to capture the desired signal because they do not have any interference cancellation processes for canceling the MAI. In addition, the blind adaptive beam forming exploiting 2-D RAKE receivers was proposed in [35–37]. Despite its capabilities of constructively combining the energy of desired signals coming from different DOAs and time delays while canceling the interference signals coming from the other DOAs, this scheme also suffers from the closely-separated-DOAs interference signals, relative to the desired signal’s DOAs, under the near-far effect situations because of insufficiency of antenna array’s degree of freedom. However, it works well in the well-separated-DOAs situation.

The interference canceling receivers using a blind array processing in the DS-CDMA system have been proposed in [19, 38], where the former proposed the optimum weight vector selection and the latter applied this receiver in the multipath environment. The iterative approach has been proposed in [19, 21], where the latter proposed to combine the parallel interference cancellation in order to cancel in-beam interference signals. One disadvantage of this iterative approach is that the user orders must be arranged from the strongest user to the weakest one. Consequently, this approach can not be suitably realized in the practical situations where the channels keep changing and the power control can not be performed perfectly. However, it works reasonably well in a situation where the DOAs are well separated and the receiver’s stages have been arranged from the strongest user down to the weakest one. Nevertheless, it does not work well in near-far effect situations. The reason lies in its optimization approach that does not consider the effect of the interfering users’ DOA situations in the system. Therefore, in order to solve such issues, the development of a new optimization approach that takes into account both the closely- and well-separated-DOA situations is needed.

A goal of this chapter is to develop a blind beam forming scheme for interference canceling receivers in the DS-CDMA systems [22] which can resolve the near-far effect situation in both the closely- and well-separated-DOA situations without using pilot signals, and relax the user-power arrangement constraint. This chapter is organized as follows. In section 2.2, a signal model of the system is described. In section 2.3, interference cancelling receivers are presented. In section 2.4, a basic blind beam forming for interference cancelling receivers in CDMA systems is reviewed. In section 2.5, a blind beam forming for interference canceling receivers, i.e. IRBAP, is proposed; and an analysis of an LCMV optimum weight vector, convergence analysis, probability of error analysis, and complexity comparison and suboptimum approach are carried out. In section 2.6, the simulation results under several situations are shown, and the concluding remark is given in section 2.7.

2.3 The Interference Canceling Receivers

A concept of interference canceling receivers [32] is employed to detect the user’s symbol as shown in Fig.2.1. The interference canceling receivers operate successively and iteratively from the first stage, which captures the strongest signal, through the Kth stage, which captures the weakest signal. After capturing the desired signal at the jth stage, the captured signal is then subtracted from the overall received signals Xj1 coming from the (j − 1)th stage. The resulting signals of this subtraction are sent to the next stages, i.e. the (j + 1)th stage and so on, consequently. The advantage of the interference cancelling receivers compared with the joint multi-user detection, e.g. an MMSE receiver and a decorrelating receiver, is its simple structure that can be easily implemented [32]. In each stage of the interference canceling receivers, the blind beamforming or IRBAP is employed as the receiver described as follows.

2.4 The Basic Blind Beamforming for Interference Canceling Receivers in DS-CDMA Systems

First of all, the basic blind beamforming without a channel equalizer for interference canceling receivers in DS-CDMA systems is reviewed. For this simple scenario, the channel is assumed static for a certain period of time, hence, the AWGN is only one source of noises and distortion in the systems. For the sake of exposition, at the first step of the state-of-the-art research, the channel gain is assumed to be unity, i.e. hk = 1, k = 1, . . . ,K. The jth stage of the basic blind beamforming without a channel equalizer for interference canceling receivers in DS-CDMA systems is illustrated in Fig.2.2. Referring to [19], the decision variable yj of the jth stage, after down-converting, sampling, and matched-filtering processes, can be expressed as follows,

Finally, in the mixed-separated-DOA situation where some of the signal’s DOAs are closely separated and the others are well separated, the IRBAP employs the interference rejection operation, by using the signature sequences of the interfering users as the key parameters for constructing the interference rejection part, for a group of signals which are closely distributed relatively to the desired signal’s DOA, meanwhile, for those signals which their DOAs are well distributed relatively to the desired signal’s DOA, the interference rejection operation is not employed. Hence, the IRBAP is able to exploit the nulling capability for nulling the well-separated-DOA interference signals while maintaining its main beam direct toward the desired signal’s DOA without being annoyed by the closely-separated-DOA interference signals; as a result, the IRBAP is more robust to the in-beam interference signals than the existing approaches in [19]. Furthermore, the average BER is significantly improved, especially in the nonordering-user-power arrangement, as will be shown in Section 2.6.

stages are set to be equal to P and the phase shifts are equal to zeros; meanwhile, the value of C’s for the remaining two stages are set to be equal to infinity and the phase shifts are calculated according to (2.57). The graph of average BER versus SNRs averaging over 100 independent runs, 10000 iterations, and K = 5 users is shown in Fig.2.10. Notice that, in the AWGN channel at BER = 10−3, the minimization method of the existing work needs to use the SNR more than the IRBAP about 10 dB, whereas the maximization method of the existing work needs about 6 dB more SNR than IRBAP. Furthermore, the IRBAP provides the BER closely to that of the theoretical BER. In the AWGN and slow-varying Rayleigh fading channels, at BER=10−3, the maximization method of the existing work needs about 3 dB more SNR than the IRBAP, whereas the minimization method of the existing work diverges at BER=10−2.

Likewise, the blind RLS based approach performs the best in the well-separated-DOAs situation; however, it suffers from the severe in-beam interference signals in the closely and mixed-separated-DOAs situations.

In the above results, it is worth noticing that the average probabilities of error of all users are greater than the results in the ordering user-power situation, especially in the AWGN and slow-varying Rayleigh fading channels. However, the proposed scheme still has lower average probabilities of error in both the AWGN channel and the AWGN and slow-varying Rayleigh fading channels. It is also worth noticing that, in high SNR regions, the theoretical BER and the BER of the IRBAP are more different than low SNR regions because the Gaussian approximation does well in the low SNR regions whereas it yields a poor approximation in the high SNR regions; and since the adaptive algorithms for adjusting weight vectors of the blind array processing and the signal canceller are employed, the detection and signal cancellation errors are inevitably occurred, especially in the transient state of the adaptation processes, resulting in higher probability of error than the theoretical result.

In Fig.2.11, the average BER versus the number of users in the nonordering-user-power mixed-separated-DOA situation, in the AWGN and slow-varying Rayleigh fading channels, at SNR = 8 dB is plotted. In addition, the number of users in the system are varying from 4 to 20 users; all users’ DOA are uniformly distributed around the 6-element ULA within [0, ]; and the user’s signal amplitude is randomly assigned in the range of [1,10]. At BER=2.2 × 10−3, it is worth noticing that the IRBAP can accommodate about 3 and 5 more users than the minimization method of the existing work and the single-element conventional receiver, respectively. Furthermore, when the number of users increases, the average BER of the IRBAP and the minimization method of the existing work gradually increase, whereas the average BER of the single-element conventional receiver rapidly increases. Obviously, the IRBAP always yields a better performance than the minimization method of the existing work and the single-element conventional receiver.

In Fig.2.12, the beam patterns of the first stage in the ordering-user-power closely separated-DOA situation at SNR = 8 dB and the DOAs of 5 users are 300, 320, 340, 360, and 380, respectively, in the AWGN and slow-varying Rayleigh fading channels are shown. The signal amplitude of 5 users are shown in table 2.1. Notice that the IRBAP and the blind RLS based approach always direct its main beam toward the first user’s DOA, whereas the minimization method of the existing work places a null toward the other users’ DOA while maintaining the main beam as close to the desired signal’s DOA as possible. Since the user arrangement is in the descending order, the minimization of the existing work is still able to preserve most of the desired signal’s strength by slightly placing the null closely to its desired signal’s DOA resulting in a slightly decrease in SINR, and hence, a slightly increase in the average BER as shown in Fig.2.5.

In Fig.2.13, the beam patterns of the first stage in the nonordering-user-power closely separated-DOA situation at SNR = 8 dB and the DOAs of 5 users are 300, 320, 340, 360, and 380, respectively, in the AWGN and slow-varying Rayleigh fading channels are shown. The signal amplitude of 5 users are shown in table 2.3. Notice that the IRBAP always directs its main beam toward the first user’s DOA, whereas the minimization method of the existing work and the blind RLS based approach place a null toward the other users’ DOA without maintaining the main beam as close to the desired signal’s DOA as possible. This phenomenon is due to the nonordering-user-power arrangement resulting in a significant decrease in SINR, and hence, a significant increase in the average BER as shown in Fig.2.8.

It is worth noticing that the results in Fig.2.12 and Fig.2.13 strongly support the derivations in Section 2.5.2.

2.7 Concluding Remark

In this chapter, the IRBAP which improves the performances of the wireless communication system in terms of the probability of error and the stability was proposed. Although the IRBAP required more complexity, which is linear with the number of users in the system, than the original scheme, the IRBAP did provide significant benefits, especially in the closely-separated-DOA situation. From the simulation results, the SNR improvement of the IRBAP over the original scheme, e.g. the minimization method, was significant in both the nonordering-user-power closely-separated-DOA and the nonordering-user-power mixed-separated-DOA situations. For instance, in the AWGN channel, the SNR difference was about 0.6 dB in the ordering-user-power mixed-separated-DOA situation, and 10 dB in the nonordering-user-power mixed-separated-DOA situation, at BER=103. Furthermore, the probability of error curves of the IRBAP were quit close to that of the theoretical BER. In the AWGN and slow-varying Rayleigh fading channels, the SNR difference was about 2 dB in the ordering-user-power mixed-separated-DOA situation, at BER=103; however, in the nonordering-user-power mixed-separated-DOA situation, the minimization method of the existing work diverged at BER=102 whereas the IRBAP still yielded a good BER.

 

CHAPTER III

A DATA-BEARING APPROACH FOR PILOT-EMBEDDING FRAMEWORKS IN SPACE-TIME CODED MIMO SYSTEMS

In the previous chapter, the smart antenna system, i.e. IRBAP, for DS-CDMA systems has been presented. In this chapter, a channel estimation for space-time coded MIMO communications systems with flat fading channels is examined. The basic background concerning the state-of-the-art channel estimation techniques is reviewed. Then, channel and system models to be considered are presented. Further, a data-bearing approach for pilot embedding frameworks, including required properties, channel estimation process, possible data bearer and pilot matrices, and data detection process, is proposed. Performance analysis as well as an optimum power allocation for the proposed scheme is also studied. Simulation results in comparison with analytical results, and concluding remark are given in the end of this chapter.

3.1 Introduction

MIMO communication systems provide prominent benefits to wireless communications due to the high capacity and reliability they can offer [2, 5]. Recently, the ST codes have been proposed in [13, 14, 52] for MIMO communications, in which the BER of the systems is significantly improved without increasing transmission power by exploiting transmit diversity [13].

A major challenge in wireless ST communications employing a coherent detector is the CSI acquisition [13, 14]. Typically, the CSI is acquired or estimated by using a pilot or training signal, a known signal transmitted from the transmitter to the receiver. This technique has been widely applied because of its feasibility for implementation with low computational complexity [16].

Two main pilot-aided channel estimation techniques have been proposed in both SISO and MIMO systems: the PSAM technique and the pilot-embedding technique. In the SISO system, the PSAM technique has been intensively studied in [16] for frequency-nonselective fading channels, and was recently extended to MIMO systems [53–58]. In this technique, firstly, a pilot signal is time-multiplexed into a transmit data stream, and then, at the receiver side, this pilot signal is extracted from the received signal to acquire the channel state information. Furthermore, an interpolation technique by averaging channel estimates over a certain time period is employed in order to improve the accuracy of the channel estimates. The disadvantage of this technique is the sparse pilot arrangement that results in poor tracking of channel variations. In addition, the denser the pilot signals, the poorer the bandwidth efficiency.

The pilot-embedding, also referred as pilot-superimposed technique, has been proposed for the SISO systems [17] and for the MIMO systems [59–61], where a sequence of pilot signals is added directly to the data stream. Some soft-decoding methods, such as Viterbi’s algorithm [17, 60], are employed for channel estimation and data detection. This technique yields better bandwidth efficiency, since it does not sacrifice any separate time slots for transmitting the pilot signal. The disadvantages of this technique lie in the higher computational complexity of the decoder and the longer delay in channel estimation process.

The purpose of this chapter is to design a novel pilot-embedding approach for ST coded MIMO systems with affordable computational cost and better fast-fading channel acquisition. The basic idea is to simplify channel estimation and data detection processes by taking advantages of the null-space and orthogonality properties of the data-bearer and pilot matrices. The data-bearer matrix is used for projecting the ST data matrix onto the orthogonal subspace of the pilot matrix. By the virtue of the null-space and orthogonality properties, in the proposed data-bearing approach for pilot-embedding, a block of data matrix is added into a block of pilot matrix, that are mutually orthogonal to each other. The benefit that is able to be expected from this approach is better channel estimation performance, since the estimator can take into account the channel variation in the transmitted data block. In addition, a low computational complexity channel estimator is also expected.

Now, the MIMO channel and system models are ready to be described. The MIMO communication system with Lt transmit antennas and Lr receive antennas is considered, as shown in Fig.3.1. In general, for a given block index t, a ST symbol matrix U(t) is an Lt × M code word matrix transmitted across the transmit antennas in M time slots. The

The rest of this chapter is organized as follows. The proposed data-bearing approach for pilot-embedding frameworks is presented in Section 3.2, including general properties needed, channel estimation process, possible data bearer and pilot matrices, and data detection process. Performance analysis for the proposed scheme is carried out in Section 3.3, in terms of channel estimation and data detection. In Section 3.4, the issue of optimum block power allocation for data and pilot parts is addressed. The simulation results are given in Section 3.5, and the chapter is concluded in Section 3.6.

This PEUB mismatch factor is used for performance measure in order to optimally allocate the powers to the data and pilot parts. In other words, this factor is minimized when the power is allocated optimally. The advantage of using this PEUB mismatch factor as a cost function for optimum power allocation inherits directly from its expression that takes both the channel estimation error and the effect of the data-bearer-projected noise into account. In addition, the use of the PEUB mismatch factor as a cost function for the optimum power allocation is better than using the channel estimation error as the cost function merely, because, under the constant power constraint, despite the fact that assigning a larger power to the pilot part yields better channel coefficient estimates, i.e. a lower channel estimation error; the remaining smaller amount of power given to the data part yields a poorer probability of error in decoding. Hence, this power tradeoff is essential for the overall performances of the pilot-embedded MIMO system, e.g. channel estimation error and the probability of detection error.

3.3.2 Nonquasi-static Flat Rayleigh Fading Channels

When the channel changes rapidly, the assumption of quasi-static fading channels is no longer held anymore. Appropriate channel estimation approaches have to be designed and analyzed for combatting such channel situations. In what follows, the performance of the proposed scheme for nonquasi-static flat Rayleigh fading channels is investigated. For the sake of exposition, a half-block fading channel model, in which the channel coefficient matrix H(t) symmetrically changes once within one ST symbol block, i.e. there exists H1(t) and H2(t) in the tth-block ST symbol matrix, is studied. With P = [P1;P2] and A = [A1;A2], the received symbol matrix in (3.9) can be rewritten as follows,

Y(t) = [H1(t)(D(t)A1 + P1);H2(t)(D(t)A2 + P2)] + N(t),                                    (3.44)

where H1(t), A1 and P1 denote the first part of the channel coefficient, the data bearer, and the pilot matrices, respectively; H2(t), A2 and P2 denote the second part of the channel coefficient, the data bearer, and the pilot matrices, respectively. In addition, the readers are reminded about the properties of matrices A and P in (3.4)-(3.7). First, the ML channel estima ion in (3.22) is computed. From the received symbol matrix in (3.44), it is post multiplied by PT , divided the result by  and rearranged the terms, the channel estimate

same MSE which coincides with the trace of the CRLB in (3.36). Notice that, the LMMSE channel estimator outperforms the ML channel estimator, where the MSE of the channel estimation is much lower in the LMMSE channel estimator. In fact, the LMMSE channel estimator is a Bayesian estimator in which the prior knowledge on the statistics of channels is exploited; therefore, its performance is much better than the ML channel estimator, which is a deterministic estimator, and that of CRLB. Furthermore, the LMMSE channel estimator tradeoffs the bias for variance, hence, the overall MSE is reduced [61]. The CRLB for Bayesian estimators including the LMMSE channel estimator can be found in [58, 61].

In Fig.3.5, we plot BERs of the pilot-embedded MIMO system with applying the optimum power allocation strategy, in comparison with the ideal-channel MIMO system, when 1 and 2-received antennas are employed. In the ideal channel case, the channel coefficients are assumed known, thus it serves as a performance bound. Notice that, at BER = 10−4, the SNR differences between the ideal-channel and the ML channel estimator are about 2.3 dB for both the 1 and 2-received antenna schemes, whereas the LMMSE channel estimation achieves the ideal-channel error probability for the 1-received antenna scheme, and the SNR difference between the ideal-channel and the LMMSE channel estimator are about 0.5 dB for the 2-received antenna scheme. In addition, the SNR differences between the ML and LMMSE channel estimators are about 1.8 dB. It is worth noticing that the LMMSE channel estimator performs better than the ML channel estimator because of the higher accurate channel estimate, as shown in Fig.3.4.

In Fig.3.6, the BERs are plotted in comparison between the proposed and alternative optimum power allocation strategies [57], both compared with the ideal-channel MIMO system, when 1 and 2-received antennas are employed. For the sake of clarity, the CMbased matrices are used as the representative of all three structures that behave similarly in the experimental results. Obviously, say at BER = 10−4, both optimum power allocation strategies are quite close resulting from the very small difference in the power allocated to the data and pilot parts in both strategies, as shown in Fig.3.3.

3.5.2 The Nonquasi-Static Flat Rayleigh Fading Channel

In this situation, the situation where the channel coefficient matrix H(t) is not kept constant over a ST symbol block is considered. An example where the channel coefficient matrix symmetrically changes twice within one ST symbol block as described in section 3.3.2 is examined. Two cases, where 1 and 2-received antennas are employed for the pilot-embedded optimum-power-allocated MIMO system, are examined in order to illustrate the effect of the time variance in the ST symbol block versus the number of received antennas.

3.5.2.1 1-Received Antenna Scheme

In Fig.3.7, the graph of MSEs of the channel estimation of the pilot-embedded MIMO system when fd ¤ T are 0.0021 (slow fading), 0.0165, 0.0412, and 0.0741 (fast fading) is shown. Notice that the CM-based matrices provides the lower MSE than the TM- and STBC-based matrices. When Doppler’s shifts are fairly large, in high SNR regimes, the SNR difference between the CM- and the TM- or STBC-based matrices ML channel estimators are approximately 6.02 dB, as remarked in the figure. This result strongly supports the derivation in section 3.3.2.1 as well.

In Fig.3.8, the graph of BERs of the pilot-embedded MIMO system when fd ¤ T are 0.0021 (slow fading), 0.0165, 0.0412, and 0.0741 (fast fading) is shown. Notice that, when Doppler’s shifts are small, e.g. fd ¤ T = 0.0021, the probability of error detection of three kinds of data bearer and pilot structures are quite the same; however, when Doppler’s shifts are getting larger, the CM-based structure is much better than the TM- and STBC-based structures, where the error floors of the CM-based structure are much lower than the TMand STBC-based structures. It is worth mentioning that, in high SNR regimes, the SNR difference between the CM-based matrices and the TM- or STBC-based matrices ML channel estimators are approximately 6.02 dB, as remarked in the figure. This result supports the derivation in section 3.3.2.2 as well. Since the nonquasi-static flat Rayleigh fading channel is the severe situation, there exists error floors that increase significantly as the Doppler’s shift increases.

3.5.2.2 2-Received Antenna Scheme

In Fig.3.9, the graph of MSEs of the channel estimation of the pilot-embedded MIMO system when fd ¤ T are 0.0021 (slow fading), 0.0412, 0.0741, and 0.1235 (fast fading) is shown. Similarly to the 1-received antenna scheme, the CM-based matrices provide the much lower MSE than the TM- and STBC-based matrices. In addition, the 6.02-dB SNR difference is also observed when Doppler’s shifts are fairly large, in high SNR regimes.

In Fig.3.10, the graph of BERs of the pilot-embedded MIMO system when fd ¤ T are 0.0021 (slow fading), 0.0412, 0.0741, and 0.1235 (fast fading) is shown. Similarly to the 1-received antenna scheme, the CM-based structure is much better than the TM- and STBC-based structures, and, in high SNR regimes, the SNR difference between the CMand the TM- or STBC-based matrices ML channel estimators are approximately 6.02 dB, as remarked in the figure.

It is worth mentioning that the CM-based structure yields better BER performances than that of the TM- and STBC-based structures, especially under the high Doppler’s shift scenarios. The reason why the CM-based structure performs better than the TM-based and STBC-based structures is that it takes both of the channel coefficient matrices H1(t) and H2(t) into account (see (3.53)), whereas the other two structures exploit either some parts of H1(t) or H2(t) based on their structures (see (3.49)). In this situation, there also exists the inevitable error floors that increase significantly as the Doppler’s shift increases. These error floors result from the channel mismatch introduced as the bias in the channel estimate, thus result in a poor detection performance especially under the high Doppler’s shift scenarios. Furthermore, the LMMSE channel estimator performs better than the ML channel estimator in low SNR regimes, in which the AWGN is the major factor that causes the detection error; however, in high SNR regimes, the channel mismatch plays a major role in causing the detection error resulting in the comparable error floors for the LMMSE and ML channel estimators. Interestingly, increasing the number of receive antennas yields an additional benefit to the ML receiver in term of the robustness to the Doppler’s shift, where the 2-received antenna scenario is less sensitive to the Doppler’s shift than the 1-received antenna scenario.

3.6 Concluding Remark

In this chapter, the data-bearing approach for pilot-embedding frameworks was proposed for joint data detection and channel estimation in ST coded MIMO systems. The main contributions of this chapter are as follows.

  • The advantages of the data-bearing approach are that it is the generalized form for pilot-embedded channel estimation and data detection in ST coded MIMO systems, in which the classical channel estimation method, e.g. PSAM, is subsumed; the low computational complexity and the efficient ML and LMMSE channel estimators are achieved; and it is capable of better acquiring the channel state information in fast-fading channels.
  • For the quasi-static flat Rayleigh fading channels, the error probability and the channel estimation performance of three data-bearer and pilot structures, i.e. the TM-, STBC-, and CM-based data-bearer and pilot matrices, are quite similar, where the optimum power-allocated schemes based on the minimum upper bound on error probability and the maximum lower bound on channel capacity optimizations yield the close results. This result claims that the proposed scheme is one of the implementable scheme that achieves the maximum lower bound on channel capacity derived in [57], in high SNR regimes. In addition, the SNR differences between the optimum-power allocated schemes and the ideal-channel schemes are about 2.3 dB when employing the unconstrained ML channel estimator and 0.5 dB for the LMMSE channel estimator.
  • For the case of nonquasi-static flat Rayleigh fading channels, the CM-based structure provide superior detection and channel estimation performances over the TM- and STBC-based structures. For instance, the 6.02 dB SNR difference is observed, as well as the error floors of the former are much smaller than that of the other two, under fairly high Doppler’s shift scenarios, in high SNR regimes.

 

CHAPTER IV

ADAPTIVE CHANNEL ESTIMATION USING PILOT-EMBEDDED DATA-BEARING APPROACH FOR SPACE-FREQUENCY CODED MIMO-OFDM SYSTEMS

In chapter III, the data-bearing approach for pilot-embedding frameworks was developed for acquiring the CSI of the frequency-nonselective (or flat) fading channels in ST coded MIMO systems. Since only a direct line-of-sight-path signal exists in such channels, a number of channel coefficients to be estimated is essentially equal to a multiplication of Lt-transmit and Lr-receive antennas, i.e. LtLr. However, in practical applications, the frequency-selective (or multipath) fading channel models are more realistic and more general than the flat fading channel models. Furthermore, the multipath fading channels are challenger than that of the flat fading channels in terms of the complication of an underlying problem, where a large number of channel coefficients are to be estimated, and the corresponding computational complexity. It is well known that the OFDM technique is one of the multicarrier modulation technique that is effectively able to combat the multipath fading channels. Recently, the OFDM technique has been introduced to MIMO systems, namely the MIMO-OFDM systems. One critical issue for such systems employing coherent receivers is channel estimation. Since the multipath delay profile of channels are arbitrary in the MIMO-OFDM systems, an effective channel estimator is needed to estimate these channels. In this chapter, the basic background about channel estimation techniques for MIMO-OFDM systems is reviewed. Further, a generalization of the pilot-embedded data-bearing approach for joint channel estimation and data detection, in which PEDB-LS channel estimation and PEDB-ML data detection are employed, is first developed. Then an LS FFT-based channel estimator by employing the concept of FFT-based channel estimation to improve the PEDB-LS channel estimation via choosing certain significant taps in constructing a channel frequency response is proposed. The effects of model mismatch error inherent in the proposed LS FFT-based channel estimator when considering non-integer multipath delay profiles, and its performance analysis are investigated. Under the framework of pilot embedding, an adaptive LS FFT-based channel estimator, that employs the optimum number of taps such that an average total energy of the channels dissipating in each tap is completely captured in order to compensate the model mismatch error as well as minimize the corresponding noise effect to improve the performance of the LS FFT-based channel estimator, is further proposed. Simulation results reveal that the adaptive LS FFT-based channel estimator is superior to the LS FFT-based and PEDB-LS channel estimators under quasi-static channels or low Doppler’s shift regimes.

4.1 Introduction

High speed data transmission services have been highly demanded in future wireless communications [67]. One promising transmission scheme to satisfy this growing demand is the use of the OFDM technique [68], in which frequency-selective fading channels are transformed into a set of parallel flat fading sub-channels. Hence, such communication techniques, e.g. channel estimation and equalization, designed for flat fading channels can be directly applied to frequency-selective fading channels through the OFDM communication scheme. In addition, since the OFDM communication scheme is a block transmission scheme, its symbol duration is longer. When the symbol duration is longer than the delay spread of channels, ISI is therefore eliminated. Nowadays, the OFDM communication scheme has been employed in various high speed wireless transmission standards such as broadband wireless LANs (IEEE 802.11a) [69], digital audio broadcasting (DAB) [70], and digital video broadcasting (DVB-T) [71]. Recently, MIMO-OFDM systems have been proposed for increasing communication capacity as well as reliability of the wireless communication systems by exploiting both the spatial and frequency diversities [68, 72]. Further the space-frequency (SF) coding for MIMO-OFDM systems have been developed for achieving such diversities in order to enhance the reception performance for high data-rate wireless communications. However, those aforementioned schemes normally need to assume an accurate CSI for coherently decoding the transmitted data, e.g. the ML decoder. Therefore, channel estimation is of critical interest for MIMO-OFDM systems.

Typically a pilot or training signal, a known signal transmitted from the transmitter to the receiver, is highly desirable to obtain an accurate channel estimation. In [73], the optimal criteria of designing the training sequence in MIMO-OFDM systems were proposed. There are two main types of pilot-aided channel estimation techniques for MIMO systems: the PSAM technique [56, 57], and the pilot-embedding technique [59, 60]. Recently, the pilot-embedded data-bearing approach for joint channel estimation and data detection has been proposed by exploiting the null-space property and the orthogonality property of the data bearer and pilot matrices [74].

Various channel estimation schemes have been recently proposed for MIMO-OFDM systems [73, 75–78]. In [75], the LMMSE channel estimator was proposed, in which SVD decomposition is used to simplify the ordinary LMMSE channel estimator. Despite the highly accurate channel estimate of this scheme, it requires intensive computational complexity and the knowledge of the underlying channel correlation. In [76], the FFT-based channel estimation using a certain number of significant taps for estimating the channel impulse response in a temporal domain was proposed. Despite the efficient computational complexity of this scheme, it could suffer from an error floor caused by a non-integer multipath delay spread, relative to the system sampling period, in the wireless channels, known as a model mismatch error. The enhancement and simplification of [76] were proposed in [77, 78], respectively.

The model mismatch error or, in the other word, the leakage effect was first mentioned in SISO-OFDM systems employing the FFT-based channel estimation [79–81]. Without the knowledge of channel correlation information, there are two ways to reduce the leakage effect: 1) by changing the exponential basis functions to the polynomial basis functions in the FFT-based channel estimation [82–84] for SISO systems and [85] for MIMO systems, and 2) by employing a proper number of taps to construct a channel frequency response in the FFT-based channel estimation [76]. In the former approach, the thorough investigation of the polynomial-based channel estimation for the MIMO systems has been conducted in [85]. Although this approach provides better performance than the FFT-based approach [81] under the non-integer multipath delay profiles, its performance is worse under the integer multipath delay profiles. Furthermore, this approach impose higher computational complexity than that of the FFT-based approach, and a general rule of designing the optimum window length as well as the optimum order of the polynomial is not fully discovered. Given the efficient implementation and reliability constraints, the FFT-based approach is still attractive. However, the optimal guideline about how to choose the number of taps remains unsolved. The challenge now is to find the optimal criteria for obtaining the optimum number of taps given that the knowledge of channel correlation information or Doppler’s shift are unavailable.

The goal of this chapter is to develop an efficient channel estimation scheme when employing pilot-embedding idea in MIMO-OFDM systems. The main contributions of this chapter are as follows.

  • A generalization of the pilot-embedded data-bearing approach for joint channel estimation and data detection for MIMO-OFDM systems, in which the PEDB-LS channel estimation and PEDB-ML data detection are employed, respectively, is developed. Furthermore, the LS FFT-based channel estimation is proposed to improve the performance of the PEDB-LS channel estimate by employing the FFT-based approach concept.
  • The model mismatch error of the LS FFT-based channel estimator is investigated, and this problem is solved by proposing an adaptive LS FFT-based channel estimation approach that employs the optimum number of taps such that the average total energy of the channels dissipating in each tap is completely captured in order to compensate the model mismatch error as well as minimize the corresponding noise effect.

The organization of this chapter is as follows. In section 4.2, the wireless channel and system models used in this chapter are introduced. In section 4.3, the generalization of the pilot-embedded data-bearing approach for joint channel estimation and data detection, including the PEDB-LS channel estimation and PEDB-ML data detection, is proposed. Under this pilot-embedding framework, in section 4.4, the LS FFT-based channel estimator is proposed, and the performance analysis for the PEDB-LS and LS FFT-based channel estimation approaches is also investigated. In section 4.5, the adaptive LS FFT-based channel estimation for improving the performance of the LS FFT-based channel estimation is proposed. In section 4.6, the performance of the proposed schemes are examined via simulations, and the conclusion is given in section 4.7.

 

4.3 Pilot-Embedded Data-Bearing Approach

In this section, the main ideas of the pilot-embedded data-bearing approach for joint channel estimation and data detection is first presented. The basic LS channel estimation and the ML data detection are then briefly introduced.

all channel estimators yields quite close results. This phenomenon stems from the fact that the channel mismatch error dominates all factors causing the detection error. In Fig.4.8, the optimum number of taps of the adaptive LS FFT-based channel estimator, when 2 receive antennas are employed, in both the quasi-static and nonquasi-static fading channels is shown. Notice that, in nonquasi-static fading channels at the particular SNR value, the number of taps is higher than the quasi-static fading channels resulting from the fact that the average energy in each PEDB-LS channel estimate tap of the former case is less than the latter because of the Doppler’s shift effect, i.e. the higher channel fluctuation in the SF-coded symbol block; as a result, in order to make the constraint of (4.39) exists, the higher number of taps is required. This phenomenon is also observed when the Doppler’s shift increases resulting in the higher number of taps for the higher Doppler’s shift case. In this scenario, there exists the inevitable error floors in the error probability, that increases significantly as the Doppler’s shift increses, resulting from the channel mismatch error introduced as the bias in the channel estimate.

4.7 Concluding Remark

In this chapter, the adaptive LS FFT-based channel estimator for improving the channel estimation and detection performances of the LS FFT-based and PEDB-LS channel estimators, and the pilot-embedded data-bearing approach for joint channel estimation and data detection were proposed. Simulations were conducted to examine the performance of the proposed schemes. For quasi-static TU-profile fading channels, the adaptive LS FFT-based channel estimator shows superior performance to that of the 10-tap LS FFTbased and PEDB-LS channel estimators. For instance, at BER of 10−4, the SNR differences are as 2.2 dB and 3.6 dB, respectively, for the adaptive LS FFT-based and the PEDB-LS channel estimators compared with the ideal-channel scheme, whereas the 10-tap LS FFTbased channel estimator suffers from the severe error floor caused by the model mismatch error. For the nonquasi-static TU-profile fading channels, under low Doppler’s shift regimes, the adaptive LS FFT-based channel estimator outperforms the 10-tap LS FFT-based and PEDB-LS channel estimators in high SNR regimes; however, in the low SNR regimes, the performance of the PEDB-LS approach is the worst and the other two estimators are comparable. Furthermore, under high Doppler’s shift regimes, the channel mismatch error dominates all factors causing the detection error and thus result in comparable error floors for all channel estimators. In addition, the LMMSE channel estimator serves as a performance bound.

CHAPTER V

CONCLUSIONS

In this dissertation, the MIMO wireless communication systems have been investigated in two aspects: the smart antenna system and channel estimation. In chapter 2, it was obvious that the smart antenna system, i.e. IRBAP, has significantly improved the performances of the interference canceling receivers in the DS-CDMA systems in terms of the probability of error and the stability of the system. Although the IRBAP requires more complexity, which is linear with the number of users in the system, than the original scheme, the IRBAP does provide significant benefits, especially in the closely-separated-DOA situation. From the simulation results, the SNR improvement of the IRBAP over the original scheme, e.g. the minimization method, is significant in both the nonordering-user-power closely-separated-DOA and the nonordering-user-power mixed-separated-DOA situations. For instance, in the AWGN channel, the SNR difference is about 0.6 dB in the ordering-user-power mixed separated-DOA situation, and 10 dB in the nonordering-user-power mixed-separated-DOA situation, at BER=103. Furthermore, the probability of error curves of the IRBAP are quit close to that of the theoretical BER. In the AWGN and slow-varying Rayleigh fading channels, the SNR difference is about 2 dB in the ordering-user-power mixed-separated-DOA situation, at BER=103; however, in the nonordering-user-power mixed-separated-DOA situation, the minimization method of the existing work diverges at BER=102 whereas the IRBAP still yields the good BER.

In chapter 3, the data-bearing approach for pilot-embedding frameworks was proposed for joint data detection and channel estimation in ST coded MIMO systems. The main contributions of this chapter are as follows.

  • The advantages of the data-bearing approach are that it is the generalized form for pilot-embedded channel estimation and data detection in ST coded MIMO systems, in which the classical channel estimation method, e.g. PSAM, is subsumed; the low computational complexity and the efficient ML and LMMSE channel estimators are achieved; and it is capable of better acquiring the channel state information in fast-fading channels.
  • For the quasi-static flat Rayleigh fading channels, the error probability and the channel estimation performance of three data-bearer and pilot structures, i.e. the TM-, STBC-, and CM-based data-bearer and pilot matrices, are quite similar, where the optimum power-allocated schemes based on the minimum upper bound on error probability and the maximum lower bound on channel capacity optimizations yield the close results. This result claims that the proposed scheme is one of the implementable scheme that achieves the maximum lower bound on channel capacity derived in [57], in high SNR regimes. In addition, the SNR differences between the optimum-power allocated schemes and the ideal-channel schemes are about 2.3 dB when employing the unconstrained ML channel estimator and 0.5 dB for the LMMSE channel estimator.
  • For the case of nonquasi-static flat Rayleigh fading channels, the CM-based structure provide superior detection and channel estimation performances over the TM- and STBC-based structures. For instance, the 6.02 dB SNR difference is observed, as well as the error floors of the former are much smaller than that of the other two, under fairly high Doppler’s shift scenarios, in high SNR regimes.

In addition, the proposed pilot-embedding scheme can be well applied to the general MIMO systems.

In chapter 4, different kinds of channel estimator have been proposed for the SF coded MIMO-OFDM systems. The main contributions of this chapter can be summarized as follows.

  • The generalization of the pilot-embedded data-bearing approach for joint channel estimation and data detection for MIMO-OFDM systems, in which PEDB-LS channel estimation and PEDB-ML data detection are employed, respectively, was developed. The LS FFT-based channel estimation was further proposed by employing the FFTbased approach concept to improve the performance of the PEDB-LS channel estimation via choosing certain significant taps in constructing a channel frequency response.
  • The performance of the LS FFT-based channel estimation was analyzed. Then the relationship between the MSE and the number of chosen taps was revealed, which in turn, the optimal criterion for choosing the optimum number of taps was explored.
  • The model mismatch error of the LS FFT-based channel estimator was investigated, and the adaptive LS FFT-based channel estimation approach was proposed to solve such problem by employing the optimum number of taps, such that the average total energy of the channels dissipating in each tap is completely captured in order to compensate the model mismatch error as well as minimize the corresponding noise effect, to construct the channel frequency response.

Simulations were conducted to examine the performance of the proposed schemes. For quasi-static TU-profile fading channels, the adaptive LS FFT-based channel estimator shows superior performance to that of the 10-tap LS FFT-based and PEDB-LS channel estimators. For instance, at BER of 10−4, the SNR differences are as 2.2 dB and 3.6 dB, respectively, for the adaptive LS FFT-based and the PEDB-LS channel estimators compared with the ideal-channel scheme, whereas the 10-tap LS FFT-based channel estimator suffers from the severe error floor caused by the model mismatch error. For the nonquasi-static TU-profile fading channels, under low Doppler’s shift regimes, the adaptive LS FFT-based channel estimator outperforms the 10-tap LS FFT-based and PEDB-LS channel estimators in high SNR regimes; however, in the low SNR regimes, the performance of the PEDB-LS approach is the worst and the other two estimators are comparable. Furthermore, under high Doppler’s shift regimes, the channel mismatch error dominates all factors causing the detection error and thus result in comparable error floors for all channel estimators. In addition, the LMMSE channel estimator serves as a performance bound. In addition, the proposed estimators can be applied to the general MIMO-OFDM systems as well.

Overall, this dissertation achieved its two objectives, which were the novel smart antenna system for DS-CDMA systems and the novel channel estimation approaches for the ST coded MIMO systems and the SF coded MIMO-OFDM systems. Firstly, it was obvious that the proposed smart antenna system, i.e. IRBAP, could be well applied to DS-CDMA  systems without requiring the descending order of power for each receiver’s stage for IRBAP. It was also robust to the near-far effect situations in such systems, and it was superior to the existing work [19] in term of the probability of detection error.

Secondly, it was clear that the proposed data bearing approach for pilot-embedding yielded significant performance improvement, including the enhanced probability of detection error in nonquasi-static fading channels, to the ST coded MIMO systems. The optimum power allocation for the pilot and the data parts for the ML channel estimator was also revealed, based on minimizing the Chernoff’s upper bound on error probability. This result claimed that this novel channel estimation approach could be applied to the real-world ST coded MIMO communications, where the channels were changing rapidly, with the enhanced probability of detection error in comparison with the traditional approach, i.e. PSAM approach.

For the SF coded MIMO-OFDM systems, the data bearing approach for pilot-embedding could be directly applied with some modifications. It was obvious that the proposed adaptive LS FFT-based channel estimator employing the optimum number of significant taps could jointly optimize the model mismatch error, caused by the non-integer multipath delay profiles for the real-world multipath fading channels, and noise effect, caused by the additive noise, inherent in the MSE of the LS FFT-based channel estimation. As a result, the enhanced probability of detection error as well as the enhanced MSE could be achieved. Hence, this channel estimator could be well applied to the real-world SF coded MIMO-OFDM communications with the enhanced performances in comparison with the proposed PEDB-LS and LS FFT-based channel estimators.

5.1 Future Works

In the future work, the extension of the IRBAP to the multipath fading channels by employing a tap delay line concept [37] is of interest. Since, here, the IRBAP regards the multipath signals as ISI, the null capability of the array antennas has been used to cancel this interference. On the other hand, the multipath signals can be exploited to improve the received SNR as well as the error probability performance of the system by coherently combining these multipath signals together. This idea resembles the concept of a RAKE receiver initially proposed for the conventional CDMA systems. This extension is straightforward, hence, it is not investigated in this dissertation. Assuming the knowledge of time delays of the multipath fading channels at the receiver, it would be interesting in exploiting the concept of the RAKE receiver to improve the performance of IRBAP.

For the data-bearing approach for pilot-embedding, an another optimum power allocation scheme based on a different criterion, e.g. the lower bound on channel capacity, is being considered for the future work. In fact, such criterion was not considered in this dissertation because this dissertation considered fixed data-rate ST coded MIMO communications. It would be of interest in comparing the performance of such systems employing different power allocation strategies.

Recently, the generalized MIMO systems (i.e. the distributed antenna array), called “Cooperative Communications,” are of widespread interest. This communications scheme can take the benefit of the MIMO system, i.e. the diversity gain, by the help of all active users in the wireless networks, e.g. the cellular phone networks. However, many issues have been waiting for being explored, including a protocol design, an effective channel estimation, synchronization, a generalized space-time code design, and scheduling and routing problems. All of these issues are of crucially interest for the future works as well.

 

References

  1. Winters, J. On the capacity of radio communication systems with diversity in a Rayleigh fading environment. IEEE Journal on Selected Areas in Communications 5 (June 1987): 871–878.
  2. Foschini, G. J. Layered space-time architecture for wireless communication in a fading environment when using multiple antennas. Bell Labs Technical Journals 1 (1996): 41–59.
  3. Foschini, G. J. and Gans, M. On limits of wireless communications in a fading environment when using multiple antennas. Wireless Personal Coomunications 6 (March 1998): 311–355.
  4. Telatar, E. Capacity of multi-antenna gaussian channels. European Transactions on Telecommunication ETT 10 (November 1999): 585–596.
  5. Telatar, E. Capacity of multi-antenna gaussian channels. AT & T Bell Labs International Memorandum (June 1995): 1–12.
  6. Goldsmith, A. Wireless Communications. UK : Cambridge University Press, 2005.
  7. Zheng, L. and Tse, D. N. Diversity and multiplexing: a fundamental tradeoff in multiple antenna channels. IEEE Transactions on Information Theory 49 (May 2003): 1073–1096.
  8. Gamal, H., Caire, G., and Damon, M. Lattice coding and decoding achieve the optimal diversity-multiplexing tradeoff of mimo channels. IEEE Transactions on Information Theory 50 (June 2004): 968–985.
  9. Yao, H. and Wornell, G. Structured space-time block codes with optimal diversitymultiplexing tradeoff and minimum delay. Proceeding of IEEE Global Telecommunications Conference 2003 (December 2003): 1941–1945.
  10. Gamal, H. E. and Damon, M. O. Universal space-time coding. IEEE Transactions on Information Theory 49, 5 (May 2003): 1097–1119.
  11. Damon, M. O., Gamal, H. E., and Beaulieu, N. C. Linear threaded algebraic space-time constellations. IEEE Transactions on Information Theory 49, 10 (October 2003): 2372–2388.
  12. Liberti, J. C. and Rappaport, T. S. Smart Antennas for Wireless Communications: IS-95 and Third Generation CDMA Applications. NJ : Prentice Hall, 1999.
  13. Tarokh, V., Seshadri, N., and Calderbank, A. R. Space-time codes for high data rate wireless communication: performance criterion and code construction. IEEE Transactions on Information Theory 44 (March 1998): 744–765.
  14. Tarokh, V., Seshadri, N., and Calderbank, A. R. Space-time block codes from orthogonal designs. IEEE Transactions on Information Theory 45 (July 1999): 1456–1467.
  15. Ding, Z. and Li, Y. (G.) Blind Equalization and Identification. NY : Marcel Dekker, 2001.
  16. Cavers, J. K. An analysis of pilot symbol assisted modulation for rayleigh fading channels. IEEE Transactions on Vehicular Technology 40 (November 1991): 686–693.
  17. Hoeher, P. and Tufvesson, F. Channel estimation with superimposed pilot sequence. Proceeding of IEEE Global Telecommunications Conference 1999 4 (December 1999): 2162–2166.
  18. Sage, A. P. and Melsa, J. L. Estimation Theory with Applications to Cimmunications and Control. NY : McGraw-Hill, 1971.
  19. Ghazi-Moghadam, V. and Kaveh, M. A cdma interference canceling receiver with an adaptive blind array. IEEE Journal on Selected Areas in Communications 16, 8 (Oct 1998): 1542–1554.
  20. Tsai, J. -A. and Woerner, B. D. Performance of combined mmse beamforming with parallel interference cancellation for overloaded ofdm-cdma systems. Proceeding of IEEE Military Communications Conference 2002 1 (October 2002): 748–752.
  21. Nazar, S. N., Ahmad, M. O., Swamy, M. N. S. and Zhu, W. -P. An adaptive parallel interference canceler using adaptive blind arrays. Proceeding of IEEE Vehicular Technology Conference 2002 1 (September 2002): 425–429.
  22. Pirak, C., Prayongpun, W., and Jitapunkul, S. The interference-rejected blind array processing for cdma interference cancellation in the near-far situation. Proceeding of World Wireless Congress 2004 (May 2004): 535–539.
  23. Hu, C. -C. and Reed, I. S. Space-time adaptive reduced-rank multistage wiener filtering for asynchronous ds-cdma. IEEE Transactions on Signal Processing 52, 7 (July 2004): 1862–1877.
  24. Al-Dabbous, N. A. and Sharif, B. S. Adaptive space-time multistage canceller for ds/cdma systems. IEE Electronics Letters 37, 6 (March 2001): 381–383.
  25. Yener, A., Yates, R. D., and Ulukus, S. Combined multiuser detection and beamforming for cdma systems: filter structures. IEEE Transactions on Vehicular Technology 51, 5 (September 2002): 1087–1095.
  26. Guo, Z. and Letaief, K. B. Adaptive mmse receiver with beamforming for ds/cdma systems. IEEE Transactions on Wireless Communications 2, 4 (July 2003): 605–610.
  27. Balanis, C. A. Antenna Theory : analysis and design. NY : Harper & Row, 1982.
  28. Keerthi, A. V., Mathur, A., Shynk, J. J., and Gooch, R. P. A blind adaptive antenna system for the estimation of mutually correlated cochannel sources. Proceeding of IEEE Military Communications Conference 1995 3 (November 1995): 1051–1055.
  29. Li, Y. and Liu, K. J. R. Adaptive blind source separation and equalization for multipleinput/multiple-output systems. IEEE Transactions on Information Theory 44, 7 (November 1998): 2864–2876.
  30. Pirak, C. and Jitapunkul, S. The JADE-estimated cm array for correlated cochannel recovery. Proceeding of IEEE Vehicular Technology Conference 2003 1 (October 2003): 1497–1501.
  31. Shynk, J. J. and Gooch, R. P. The constant modulus array for cochannel signal copy and direction finding. IEEE Transactions on Acoustics, Speech, and Signal Processing 44, 3 (March 1996): 652–660.
  32. Verdu, S. Multiuser Detection. UK. : Cambridge University Press, 1998.
  33. Dominique, F. and Reed, J. H. Despread data rate update multitarget adaptive array for cdma signals. IEE Electronics Letters 33, 2 (January 1997): 119–121.
  34. Rong, Z., Petrus, P., Rappaport, T. S., and Reed, J. H. Despread-respread multi-target constant modulus array for cdma systems. IEEE Communications Letters 1, 4 (July 1997): 114–116.
  35. Zoltowski, M. D. and Ramos, J. Blind adaptive beamforming for cdma based pcs/cellular. Proceeding of IEEE Asilomar Conference on Signals, Systems and Computers 1995 (November 1995): 378–382.
  36. Zoltowski, M. D. and Ramos, J. Blind multi-user access interference cancellation for cdma based pcs/cellular using antenna arrays. Proceeding of IEEE International Conference on Acoustics, Speech, and Signal Processing 1996 (May 1996): 2730–2733.
  37. Chen, Y. -F. and Zoltowski, M. D. Blind rls based space-time adaptive 2-d rake receivers for ds-cdma communication systems. IEEE Transactions on Signal Processing 48, 7 (July 2000): 2145–2150.
  38. Vaidynathan, C., Buckley, K. M., and Hosur, S. A. Blind adaptive antenna array for cdma systems. Proceeding of IEEE Asilomar Conference on Signals, Systems and Computers 1995 2 (November 1995): 1373–1377.
  39. Veen, A.-J. van der Algebraic methods for deterministic blind beamforming. IEEE Proceedings 86 (October 1998): 1987–2008.
  40. Sidiropoulos, N. D., Giannakis, G. B., and Bro, R. Blind parafac receivers for ds-cdma systems. IEEE Transactions on Signal Processing 48, 3 (March 2000): 810–823.
  41. Weiss, A. J. and Friedlander, B. Channel estimation for ds-cdma downlink with a periodic spreading codes. IEEE Transactions on Communications 47, 10 (October 1999): 1561–1569.
  42. Li, K. and Liu, H. Channel estimation for ds-cdma with aperiodic spreading codes. Proceeding of IEEE International Conference on Acoustics, Speech, and Signal Processing 1999 5 (March 1999): 2535–2538.
  43. Monzingo, R. A. and Miller, T. W. Introduction to Adaptive Arrays. NY : Wiley, 1980.
  44. Shahbazpanahi, S., Gershman, A. B., Luo, Z. -Q., and Wong, K. M. Robust adaptive beamforming for general-rank signal models. IEEE Transactions on Signal Processing 51, 9 (September 2003): 2257–2269.
  45. Vorobyov, S. A., Gershman, A. B., and Luo, Z. -Q. Robust adaptive beamforming using worst-case performance optimization: a solution to the signal mismatch problem. IEEE Transactions on Signal Processing 51, 2 (February 2003): 313–324.
  46. Li, J., Storica, P., and Wang, Z. On robust capon beamforming and diagonal loading. IEEE Transactions on Signal Processing 51, 7 (July 2003): 1702–1715.
  47. Lee, C. -C. and Lee, J. -H. Robust adaptive array beamforming under steering vector errors. IEEE Transactions on Antennas Propagation 45, 1 (January 1997): 168–175.
  48. Cox, H., Zeskind, R. M., and Owen, M. M. Robust adaptive beamforming. IEEE Transactions on Acoustics, Speech, and Signal Processing ASSP-35, 10 (October 1987): 1365–1376.
  49. Haykin, S. Adaptive Filter Theory. NJ : Prentice-Hall International, 1996.
  50. Proakis, J. G. Digital Communications, 3rd MA : McGraw-Hill, 2001.
  51. Jakes, Jr. W. C. Multipath Interference. NY : Microwave Mobile Communication, Wiley, 1974.
  52. Alamouti, S. M. A simple transmitter diversity scheme for wireless communications. IEEE Journal on Selected Areas in Communications 16 (Oct 1998): 1451–1458.
  53. Cheon, H. and Hong, D. Performance analysis of space-time block codes in time-varying rayleigh fading channels. Proceeding of IEEE International Conference on Acoustics, Speech, and Signal Processing 2002 3 (May 2002): 2357–2360.
  54. Kotecha, J. H., and Sayeed, A. M. Transmit signal design for optimal estimation of correlated mimo channels. IEEE Transactions on Signal Processing 52 (February 2004): 546–557.
  55. Samardzija, D. and Mandayam, N. Pilot-assisted estimation of mimo fading channel response and achievable data rates. IEEE Transactions on Signal Processing 51 (November 2003): 2882–2890.
  56. Song, Y. and Blostein, S. D. Channel estimation and data detection for mimo systems under spatially and temporally colored interference. EURASIP Journals on Applied Signal Processing (May 2004): 685–695.
  57. Hassibi, B. and Hochwald, B. M. How much training is needed in multiple-antenna wireless links?. IEEE Transactions on Information Theory 49 (April 2003): 951–963.
  58. Dong, M. and Tong, L. Optimal design and placement of pilot symbols for channel estimation. IEEE Transactions on Signal Processing 50 (December 2002): 3055–3069.
  59. Budianu, C. and Tong, L. Channel estimation for space-time orthogonal block codes. IEEE Transactions on Signal Processing 50 (October 2002): 2515–2528.
  60. Zhu, H., Farhang-Boroujeny, B., and Schlegel, C. Pilot embedding for joint channel estimation and data detection in mimo communication systems. IEEE Communications Letters 7 (January 2003): 30–32.
  61. Vosoughi, A. and Scaglione, A. Everything you wanted to know about training: guidelines derived using the affine precoding framework and the crb. to appear in IEEE Transactions on Signal Processing.
  62. Pease, M. C. Methods of Matrix Algebra. NY : Academic Press, 1965.
  63. Perlis, S. Theory of Matrices. MA : Addison-Wesley Press, 1952.
  64. Geramita, A. V. and Seberry, J. Orthogonal Designs. NY : Marcel Dekker, 1979.
  65. Khuri, A. I. Advanced Calculus with Applications in Statistics. NY : Wiley, 1993.
  66. Tarokh, V., Seshadri, N., and Calderbank, A. R. Space-time block coding for wireless communications: performance results. IEEE Journal on Selected Areas in Communications 17 (March 1999): 451–460.
  67. Ojanper¨a, T. and Prasad, R. An overview of third generation wireless personal communications: A European perspective. IEEE Personal Communications 5 (December 1998): 59–65.
  68. Cimini, Jr. L. J. and Sollenberger, N. R. Ofdm with diversity and coding for high-bitrate mobile data applications. Proceeding of IEEE Global Telecommunications Conference 1997 (November 1997): 305–309.
  69. IEEE, IEEE Std 802.11a: Wireless LAN Medium Access Control (MAC) and physical layer (PHY) specifications: High-speed physical layers in the 5GHz band, 1999.
  70. ETS, ETS Std 300 401: Radio Broadcasting Systems; Digital Audio Braodcasting (DAB) to mobile, portable and fixed receivers, 1997.
  71. ETS, ETS Std 300 744, v1.2.1: Digital Video Broadcasting (DVB) framing structure, channel coding and modulation digital terestrail television (DVB-T), 1999.
  72. Su, W., Safar, Z., Olfat, M., and Liu, K. J. R. Obtaining full-diversity space-frequency codes from space-time codes via mapping. IEEE Transactions on Signal Processing 51, 11 (November 2003): 2905–2916.
  73. Barhumi, I., Leus, G., and Moonen, M. Optimal training design for mimo-ofdm systems in mobile wireless channels. IEEE Transactions on Signal Processing 51, 6 (June 2003): 1615–1624.
  74. Pirak, C., Wang, Z. J., Liu, K. J. R., and Jitapunkul, S. Performance analysis for pilotembedded data-bearing approach in space-time coded mimo systems. Proceeding of IEEE International Conference on Acoustics, Speech, and Signal Processing 2005 (March 2005): 593–596.
  75. Gong, Y. and Lataief, K. B. Low complexity channel estimation for space-time coded wideband ofdm systems. IEEE Transactions on Wireless Communications 2, 5 (September 2003): 876–882.
  76. Li, Y. (G.) Channel estimation for ofdm systems with trsmitter diversity in mobile wireless channels. IEEE Journal on Selected Areas in Communications 17, 3 (March 1999): 461–471.
  77. Li, Y. (G.), Winters, J. H., and Sollenberger, N. R. Mimo-ofdm for wireless communications: signal detection with enhanced channel estimation. IEEE Transactions on Communications 50, 9 (September 2002): 1471–1477.
  78. Li, Y. (G.) Simplified channel estimation for ofdm systems with multiple transmit antennas. IEEE Transactions on Wireless Communications 1, 1 (January 2002): 67–75.
  79. van de Beek, J. J., Edfors, O., Sandell, M., Wilson, S. K., and Borjesson, P. O. On channel estimation in ofdm systems. Proceeding of IEEE Vehicular Technology Conference 1999 (July 1999): 815–819.
  80. Edfors, O., van de Beek, J. J., Wilson, S. K., Sandell, M., and Borjesson, P. O. Ofdm channel estimation by singular value decomposition. IEEE Transactions on Communications 46 (July 1998): 931–939.
  81. Li, Y. (G.), Cimini, L. J., and Sollenberger, N. R. Robust channel estimation for ofdm systems with rapid dispersive fading channels. IEEE Transactions on Communications 46 (July 1998): 902–915.
  82. Chang, M, -X., and Su, Y. T. Model-based channel estimation for ofdm signals in Rayleigh fading. IEEE Transactions on Communications 50, 4 (April 2002): 540–544.
  83. Wang, X. and Liu, K. J. R. Channel estimation for multicarrier modulation systems using a time-frequency polynomial model. IEEE Transactions on Communications 50, 7 (July 2002): 1045–1048.
  84. Pandana, C., Sun, Y., and Liu, K. J. R. Channel-aware priority transmission scheme using joint channel estimation and data loading for ofdm systems. to appear in IEEE Transactions on Signal Processing.
  85. Wang, X. and Liu, K. J. R. Model based channel estimation framework for mimo multicarrier communication systems. IEEE Transactions on Wireless Communications 4, 3 (May 2005): 1–14.
  86. Steele, R. Mobile Radio Coomunications. NY : IEEE Press, 1992.
  87. Stuber, G. Principles of Mobile Communications. MA : Kluwer, 2001.

 

Appendices

Appendix A

List of Abbreviations

A/D    –                       analog to digital

AWGN           –           additive white Gaussian noise

BER     –                       bit error rate

bps     –                       bit per second

CDMA           –           code-division multiple access

CM     –                       code multiplexing

CMA –                       constant modulus algorithm

CRLB –                       Cramer-Rao lower bound

CSI      –                       channel state information

D/A    –                       digital to analog

DAB    –                       digital audio broadcasting

DC      –                       down converting

DD      –                       decision-directed

DFT     –                       discrete Fourier transform

DOA   –                       directions of arrival

DR       –                       despread-respread

DS-CDMA     –           direct sequence code division multiple access

DSL     –                       digital subscriber lines

DVB-T             –           digital video broadcasting

FCC    –                      federal communications commission

FDMA            –           frequency-division multiple access

FIR       –                       finite impulse response

FFT      –                       fast Fourier transform

HF       –                       high frequency

HP       –                       half-width

IDFT    –                       inverse discrete Fourier transform

IFFT     –                       inverse fast Fourier transform

i.i.d.     –                       independent and identically distributed

IRBAP             –           interference-rejected blind array processing

ISI        –                       intersymbol interference

ISM     –                       industrial, scientific, and medical

LANs   –                       local area networks

LCMV            –           linear constraint minimum variance distortionless response

LMMSE           –           linear minimum mean square error

LMS     –                       least mean square

LS        –                       least squares

MAI    –                       multiple access interference

Max    –                       SNR maximum signal-to-noise ratio

MIMO            –           multiple-input multiple-output

ML       –                       maximum likelihood

MMSE            –           minimum mean square error

MPSK –                       M-phased-shift-keyed

MSE    –                       mean square error

MTSO             –          mobile telephone switching office

OFDM            –           orthogonal frequency division multiplexing

OFDMA         –           orthogonal frequency-division multiple access

PEDB             –           pilot-embedded data-bearing

PSAM             –           pilot-symbol assisted modulation

PSK                 –           phase shift keying modulation

QAM             –           quadrature amplitude modulation

RLS                  –           recursive least square

SF                    –           space-frequency

SISO                –           single-input single-output

SNR                 –           signal-to-noise ratio

SP                    –           Sampling

ST                    –           space-time

STBC               –           space-time block code

SVD                –           singular value decomposition

TDMA             –           time-division multiple access

TM                   –           time multiplexing

TU                    –           typical urban

ULA                 –           uniform linear array

U-NII               –           unlicensed national information infrastructure

W-CDMA      –           wideband code-division multiple access

ZF                    –           zero-forcing

ZMSW             –           zero-mean spatially white

Leave a Comment Here